WO2003038806A1 - Procedes et appareil de determination de la hauteur tonale - Google Patents

Procedes et appareil de determination de la hauteur tonale Download PDF

Info

Publication number
WO2003038806A1
WO2003038806A1 PCT/US2002/033895 US0233895W WO03038806A1 WO 2003038806 A1 WO2003038806 A1 WO 2003038806A1 US 0233895 W US0233895 W US 0233895W WO 03038806 A1 WO03038806 A1 WO 03038806A1
Authority
WO
WIPO (PCT)
Prior art keywords
vectors
pairs
sequence
signal
histogram
Prior art date
Application number
PCT/US2002/033895
Other languages
English (en)
Inventor
Dmitry Edward Terez
Original Assignee
Dmitry Edward Terez
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Family has litigation
First worldwide family litigation filed litigation Critical https://patents.darts-ip.com/?family=26837975&utm_source=google_patent&utm_medium=platform_link&utm_campaign=public_patent_search&patent=WO2003038806(A1) "Global patent litigation dataset” by Darts-ip is licensed under a Creative Commons Attribution 4.0 International License.
Application filed by Dmitry Edward Terez filed Critical Dmitry Edward Terez
Publication of WO2003038806A1 publication Critical patent/WO2003038806A1/fr

Links

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/90—Pitch determination of speech signals

Definitions

  • the present invention relates generally to a signal processing and, more particularly, to methods and apparatus for detecting periodicity and/or for determining the fundamental frequency of a signal, for example, a speech signal.
  • a problem frequently encountered in many signal processing applications is to determine whether a portion of a signal is periodic or aperiodic and, in case it is found to be periodic, to measure the period length. This task is particularly important in processing acoustic signals, like human speech or music.
  • the term "pitch" is used to refer to a fundamental frequency of a periodic or quasi-periodic signal.
  • the fundamental frequency may be, e.g., a frequency, which may be perceived as a distinct tone by the human auditory system.
  • Fundamental frequency is defined as the inverse of the fundamental period for some portion of a signal.
  • Pitch in human speech is manifested by nearly repeating waveforms in periodic "voiced" portions of speech signals, and the period between these repeating waveforms defines the pitch period.
  • voiced speech sounds are produced by periodic oscillations of human vocal cords, which provide a source of periodic excitation for the vocal tract.
  • Unvoiced portions of speech signals are produced by other, non-periodic, sources of excitation and normally do not exhibit any periodicity in a signal waveform.
  • most of the conventional short-term pitch-determination methods belong to one of the following three groups: (1) methods based on auto- or cross-correlation of a signal, (2) frequency-domain methods analyzing harmonic structure of a signal spectrum and (3) methods based on cepstrum calculation.
  • correlation-based pitch determination has one major drawback - the presence of secondary peaks due to speech formants (vocal tract resonances), in addition to main peaks corresponding to pitch period and its multiples. This property of the correlation function makes the selection of correct peaks very difficult. In order to circumvent this difficulty some sophisticated post-processing techniques, like dynamic programming, are commonly used to select proper peaks from computed correlation functions and to produce correct pitch contours.
  • Cepstrum-based methods are not particularly sensitive to speech formants, but tend to be rather sensitive to noise.
  • a cepstrum-based approach lacks generality: it fails for some simple periodic signals.
  • a cepstrum-based approach is unable to determine the fundamental period of an extremely band-limited signal, such as pure sine wave.
  • cepstrum-based pitch detectors would fail in such instances, i.e., they would fail on an otherwise clearly periodic signal with a well- defined pitch.
  • frequency-domain pitch-determination methods run into difficulties when the fundamental frequency component is actually missing in a signal, which is often the case with telephone-quality speech signals.
  • Speech generation by a human vocal apparatus is a very complex nonlinear and non-stationary process, of which there is only an incomplete understanding.
  • To achieve a complete and precise understanding of human speech production it needs to be described in terms of nonlinear fluid dynamics.
  • this kind of description cannot be used directly for building signal processing devices.
  • speech production has been described in terms of a source-filter model, which gives a good approximation for many purposes, but is inherently limited in its ability to model the true dynamics of speech production.
  • the present invention is directed to methods and apparatus for pitch and periodicity determination in speech and/or other signals. It is also directed to methods and apparatus for pitch tracking and/or for detecting voiced or unvoiced portions in speech signals.
  • information about pitch and periodicity of a signal is obtained using methods of signal embedding into a multi-dimensional state space, originally introduced in the theory of nonlinear and chaotic signals and systems.
  • speech signal is acquired and pre-processed in a known manner, by performing processing including analog-to-digital conversion.
  • a sampled digitized signal is represented, in a conventional way, as a sequence of frames, each frame including a predetermined number of samples.
  • Each frame is embedded into an m- dimensional state space by using an embedding procedure.
  • a time-delay embedding procedure is used with a fixed embedding dimension, e.g., of three, and a constant delay parameter equal to a predetermined number of samples. This embedding procedure transforms each frame into a sequence of m-dimensional vectors describing a trajectory in m-dimensional state space.
  • closest pairs of vectors are selected from a plurality of possible pairs of vectors in the sequence of m-dimensional vectors. Closest pairs of vectors represent nearest-neighbor points on the reconstructed trajectory and have the smallest distances between vectors in m-dimensional state space. Euclidean distances in m- dimensional space are used in the aforementioned exemplary embodiment, but other distance norms can also be used. In one embodiment, closest pairs of vectors are selected by identifying pairs of vectors with a distance between vectors in state space less than a predetermined, e.g., set, neighborhood radius. Each pair of vectors has a certain time separation between vectors which can be expressed in terms of a number of samples.
  • a periodicity histogram is obtained by accumulating total numbers of the selected closest pairs of vectors with the same time separations between vectors in corresponding histogram bins.
  • the obtained histogram is characterized by distinct peaks corresponding to a fundamental period and its integer multiples for periodic signals, and by the absence of such peaks for non-periodic signals.
  • Each bin in the periodicity histogram can be normalized with respect to its maximal possible value to obtain a normalized periodicity histogram.
  • the periodicity histogram generated in accordance with the invention is a function of a number of selected closest pairs, or equivalently, of a chosen neighborhood radius in state space.
  • a reconstructed trajectory for each frame is normalized to fit into a unit cube in state space, and a constant predetermined neighborhood radius is used for selecting closest pairs of vectors.
  • an adaptive procedure for selecting an appropriate number of closest pairs is used. The adaptive procedure performs selection of the closest pairs based on the detected magnitude of the highest histogram peak, in order to make main histogram peaks more reliable and easy to identify.
  • the obtained periodicity histogram is searched for highest peaks in a predetermined interval of possible pitch values.
  • the position of the highest peak in the periodicity histogram is used as a local estimate of the pitch period in samples.
  • a normalized periodicity histogram is used to identify one or more highest peaks, and the positions of the identified peaks are then used as pitch period candidates for further post-processing.
  • a post-processing technique can be, and in various embodiments is, employed to construct a pitch track and to perform voiced/unvoiced segmentation of a speech signal.
  • Various suitable post-processing methods e.g. dynamic programming, can be used with the present invention.
  • One feature of the present invention is directed to a simple and efficient method for performing simultaneous pitch tracking and voiced/unvoiced segmentation of speech signals with minimal processing delay.
  • speech frames are classified as either "reliable” or "unreliable".
  • a speech frame is classified as reliable, if it has one or more pitch period candidates and, in case of several pitch candidates, they are integer multiples of the lowest candidate's value. Additional conditions can also be imposed to determine if the frame is reliable. Other frames, e.g., all other frames in one embodiment, are classified as unreliable.
  • a start of voicing determination is made when a sequence of several (two in one particular exemplary embodiment) consecutive reliable frames is encountered, provided that their corresponding pitch candidates match each other. After the start of a voiced segment is determined, a pitch-tracking procedure attempts to track pitch period backward and forward in time.
  • the maximal number of frames to track backward may be limited by the maximal allowed processing delay.
  • the pitch-tracking procedure searches a plurality of pitch candidates for the best match to the current pitch estimate, subject to constraints of pitch continuity for consecutive voiced frames. When the pitch track can no longer be continued, an unvoiced decision is made.
  • alternative embedding procedures can be used in place of time-delay embedding.
  • One particular alternative embedding procedure is singular value decomposition embedding, which can be advantageous for noisy signals.
  • a method of forming pairs of vectors for selecting the closest pairs can be modified, in order to have the same maximal value for each histogram bin.
  • FIG. 1A illustrates a speech frame of 220 samples of speech corresponding to the sustained vowel /AA/.
  • FIG. IB illustrates time-delay embedding in 3-dimensional state space of the speech frame illustrated in FIG. 1A.
  • FIG. 2A illustrates a space-time separation plot for the embedded speech frame illustrated in FIG. IB.
  • FIG. 2C illustrates a normalized periodicity histogram obtained from the histogram illustrated in FIG. 2B.
  • FIG. 3C illustrates an unbiased auto-correlation function computed for the speech frame illustrated in FIG. 1 A.
  • FIG. 4A illustrates a speech frame of 220 samples of the transitional voiced segment of speech.
  • FIG. 4B illustrates time-delay embedding in 3-dimensional state space of the speech frame illustrated in FIG. 4A.
  • FIG. 5A illustrates a space-time separation plot for the embedded speech frame illustrated in FIG. 4B.
  • FIG. 5C illustrates a normalized periodicity histogram obtained from the histogram illustrated in FIG. 5B.
  • FIG. 6C illustrates an unbiased auto-correlation function computed for the speech frame illustrated in FIG. 4A.
  • FIG. 7A illustrates a speech frame of 220 samples of the fricative /S/.
  • FIG. 7B illustrates time-delay embedding in 3-dimensional state space of the speech frame illustrated in FIG. 7A.
  • FIG. 8A illustrates a space-time separation plot for the embedded speech frame illustrated in FIG. 7B.
  • FIG. 8C illusfrates a normalized periodicity histogram obtained from the histogram illustrated in FIG. 8B.
  • FIG. 9C illustrates an unbiased auto-correlation function computed for the speech frame illustrated in FIG. 7A.
  • FIG. 10 is a flowchart illustrating the basic steps involved in determining pitch in accordance with the present invention.
  • FIG. 11 is a flowchart illusfrating an adaptive method of selecting closest pairs of vectors for a periodicity histogram in accordance with one embodiment of the invention.
  • FIG. 12 is a flowchart of the pitch-tracking method according to one particular embodiment of the invention.
  • FIG. 13 A illustrates a speech signal waveform for the male-spoken utterance "She had your dark suit” sampled at 16 kHz.
  • FIG. 13B illustrates fundamental frequency contours obtained with the method of the present invention for the speech signal waveform illustrated in FIG. 13 A.
  • FIGS. 14 A, 14B and 14C illustrate results of an SVD-embedding for the speech frames illustrated in FIGS. 1 A, 4 A and 7 A, respectively.
  • FIGS. 15A illusfrates a method of generating all possible pairs of vectors for selecting the closest pairs according to one exemplary embodiment of the invention.
  • FIGS. 15B illustrates a method of generating a subset of all possible pairs of vectors for selecting the closest pairs in accordance with one alternative embodiment of the invention.
  • FIG. 16 is a schematic block diagram of a pitch-determination apparatus in accordance with the present invention.
  • Human speech is generated by a highly complex nonlinear dynamical system, yet the only observable output of this system for most practical purposes is a speech signal. Accordingly, a scalar one-dimensional speech signal can be used to reconstruct a multi-dimensional state space topologically equivalent to the original state space, in which the complex nonlinear dynamics of human speech production take place.
  • Signal Embedding
  • Processing speech or any other signal in accordance with the present invention begins with signal embedding into an m-dimensional state space. This step is normally preceded by a signal pre-processing stage, which may be implemented using known techniques. Preprocessing normally includes analog-to-digital conversion that produces a sampled digitized signal. For example, in one particular embodiment of the invention, a speech signal is sampled at 16 kHz with 16-bit linear-scale accuracy. Some optional signal conditioning can also be applied to a signal in the pre-processing stage.
  • the method of the present invention can work on raw digitized speech signals and does not explicitly require any signal pre-conditioning. However, in many cases using some conventional signal-conditioning techniques, like moderate low-pass filtering, can improve the quality of results.
  • a sampled digitized signal is represented, in a usual way, as a sequence of (overlapping) frames.
  • Each frame includes a portion of the sampled digitized signal, or a sequence of successive samples.
  • each frame includes a constant number of samples N.
  • each frame usually include at least two complete pitch periods.
  • One of the important advantages of the present invention is that it can produce reliable pitch estimates with frames shorter than two (but longer than one) complete pitch periods in the case of clean periodic signals.
  • the upper limit on a frame size is dictated by a range of possible pitch periods and by resolution requirements.
  • N should preferably be chosen such that each frame does not include too many pitch periods.
  • This value of N can be used for most female voices (with F0 in the range 100 - 400 Hz, for example), provided that speech signal is clean and sampled at 16 kHz. For other voices and sampling rates the value of N should be chosen appropriately.
  • Variable-sized frames can also be used in other embodiments of the invention.
  • a sampled signal in each frame is embedded into m-dimensional state-space by use of an embedding procedure.
  • the embedding procedure used in the exemplary embodiment is time-delay embedding.
  • vectors x(i) in m-dimensional state space are formed from time-delayed values of a signal s(i):
  • m is the embedding dimension and d is the delay parameter, or lag (in integer number of samples).
  • m- dimensional vector and "point in m-dimensional space” have the same meaning in this description: a set of m independent coordinates uniquely defining location in m-dimensional space).
  • These m-dimensional vectors x(i) correspond to successive points on a reconstructed frajectory in m-dimensional state space, which is topologically equivalent to the original state space of a signal-generating system, e.g., a nonlinear speech generation process.
  • the rows contain m- dimensional vectors x(i) describing the frajectory in m-dimensional state space reconstructed using time-delay embedding.
  • the reconstructed frajectory for a steady periodic signal such as sustained vowel in FIG. 1A, has a clear periodic nature. Note that the frajectory in FIG. IB almost repeats itself after a complete pitch period. This periodicity is less evident in the state-space reconstruction of the transitional voiced segment, such as the one shown in FIG. 4B. For the unvoiced aperiodic fricative, the reconstructed vectors tend to randomly fill the state space, as illustrated in FIG. 7B.
  • voiced speech sounds can be sufficiently embedded in 3- dimensional state space, whereas unvoiced speech sounds (e.g. fricatives) have a high- dimensional nature.
  • the optimal value of the delay parameter d in an integer number of samples depends on the sampling rate and on signal properties.
  • the delay parameter should be large enough for a reconstructed frajectory of each frame to be sufficiently "open" in state space. On the other hand, it is desirable to keep the delay parameter relatively small for better resolution.
  • a constant delay parameter d is used for embedding all frames.
  • delay parameter d may be chosen differently or even determined independently for each speech frame, in order to adapt to signal properties. It should be noted that the actual mode of implementing time-delay embedding in accordance with EQ. 1 can differ in various embodiments of the invention.
  • a sampled digitized signal is segmented into short (overlapping) frames of N samples each, as discussed above, and each frame is independently embedded according to EQ. 2.
  • Euclidean distance norm in m-dimensional space may be used as a spatial distance:
  • the squared Euclidean distances are used to reduce computations when computing and comparing distances in the exemplary embodiment.
  • the use of squared distances avoids the need to perform square root computations.
  • Distance norms in m-dimensional space other than Euclidean can, and in some embodiments are, used in alternative embodiments of the invention.
  • one-norm is used in one alternative embodiment:
  • distances can be measured relative to the maximal size of the reconstructed frajectory in state space.
  • a reconstructed frajectory for each frame is normalized to fit into the unit cube in m-dimensional state space. This normalization can be achieved by linear scaling and shifting of each dimension, so that each dimension of the trajectory is between 0 and 1.
  • each dimension of the frajectory, reconstructed using time-delay embedding is a delayed version of the same signal, similar normalization can be achieved by normalizing a sequence of samples in each individual frame prior to time-delay embedding.
  • a useful graphical tool for visualizing a distribution of spatial distances and time separations between vectors on the reconstructed trajectory is a space-time separation plot, originally introduced by Provenzale, A. et al. for qualitative analysis of chaotic time-series ("Distinguishing between low-dimensional dynamics and randomness in measured time series", Physica D 58, 1992, pp. 31-49). It is a simple scatter plot of spatial distance D[x(i),x(j)] versus time separation ⁇ i-j ⁇ for each possible pair of vectors ⁇ x(t), x(j) ⁇ on the frajectory. It should be understood that a space-time separation plot is not needed to practice the invention. Rather, it is used to provide a graphical illustration of basic concepts.
  • FIGS. 2 A, 5 A and 8A show space-time separation plots for the reconstructed trajectories of a sustained vowel /AA/, a transitional voiced segment and a fricative /S/, each of which is illustrated in FIGS. IB, 4B and 7B, respectively. Only the lower parts of the entire plots are actually shown.
  • FIG. 2A shows that, in the case of a periodic vowel, data points with small spatial distances tend to concentrate around time separation values corresponding to a fundamental pitch period and its integer multiples.
  • For a transitional voiced segment some vertical regions of data point concentration are also clearly visible in FIG. 5A.
  • the unvoiced fricative /S/ data points in the space-time separation plot are randomly distributed along a time separation axis, as it is evidenced by FIG. 8A.
  • Closest pairs of vectors are pairs of vectors ⁇ x( , x(/) ⁇ wi h the smallest spatial distances D[x(i),x(j)] between vectors among possible pairs of vectors in the sequence of m-dimensional vectors x(t) ( ⁇ — 1...M).
  • the computed distances are then compared with the predetermined value of r, and pairs with a distance D[x( ⁇ ),x(j)) ⁇ r are selected as closest pairs.
  • squared Euclidean distances are computed.
  • the computed distances are compared with the squared value of r.
  • closest pairs of vectors are selected by computing spatial distances D[x(i),x(j)] for all possible non-repeating pairs of vectors in the sequence of vectors x(t) (/-l ...M), ordering vector pairs by their spatial distances in increasing order, and selecting a predetermined number n of closest pairs from the ordered set of vector pairs. The selection can be easily performed as a result of the ordering.
  • a periodicity histogram is computed based on time separation values of the selected closest pairs of vectors.
  • Each bin in the periodicity histogram accumulates a total number of selected closest pairs having the same time separation between vectors, e.g., as expressed by the number of samples corresponding to a bin index.
  • the term "histogram" in this description is used to refer to a one-dimensional array of numbers, where each bin in a histogram corresponds to an element of the one-dimensional array.
  • Periodicity histogram computation can be performed by summing up data points with the same horizontal positions (that is, lined up vertically) and located below line 22 in the space-time separation plot of FIG. 2A, to yield the histogram shown in FIG. 2B.
  • a periodicity histogram For the sequence of vectors x(i) (i-l ...M) representing a frajectory in m- dimensional state space, a periodicity histogram can be formally defined as
  • M-k hist(k) Y H(r -D[x(i),x(i + k)]) EQ. 4
  • k is a bin index corresponding to the time separation in samples between vectors x( ⁇ ) and x(i+k)
  • r is a predetermined neighborhood radius
  • D[x(i),x(i+k)] is a spatial distance between vectors
  • H is ⁇ eaviside function.
  • Euclidean spatial distance between vectors used in the exemplary embodiment, can be replaced with some other distance norm in m-dimensional space.
  • FIGS. 2B, 5B and 8B show periodicity histograms computed according to EQ. 4 with r — 0.1 for a sustained vowel I ⁇ J, a transitional voiced segment and an unvoiced fricative /S/, respectively.
  • FIG. 2B shows a sharp peak 24 corresponding to the fundamental pitch period of a periodic vowel, and a second sharp peak 26 corresponding to twice the pitch period value.
  • the periodicity histogram in FIG. 5B, computed for the transitional voiced segment shows a peak 52 corresponding to a fundamental pitch period. However, in this case the peak 52 is much lower and is not sharp.
  • the periodicity histogram for the unvoiced fricative /SI in FIG. 8B shows many random low peaks distributed along the time separation axis.
  • a periodicity histogram computed according to EQ. 4 with an appropriately chosen value of r (or equivalently, with an appropriate number of selected closest pairs of vectors), will have distinct peaks corresponding to a fundamental period and its integer multiples for periodic signals. Periodicity histograms corresponding to aperiodic signals will lack such characteristic peaks.
  • a periodicity histogram Since the summation interval in EQ. 4 linearly shrinks with an increasing value of A:, a periodicity histogram has a bias: an upper bound is not the same for all bins and is a linearly decaying function of k, as shown by slanting line 28 in FIG. 2B. This causes the magnitudes of histogram peaks to decay with increasing values of k, as it is observed in FIG. 2B. Due to this decay, the main histogram peak, corresponding to the lowest sub-multiple and representing a true fundamental period, is usually the largest of all peaks for clean and steady periodic signals, as it is evidenced by peak 24 in FIG. 2B. Thus, locating the highest peak in the periodicity histogram can give a reliable pitch period estimate for clean and steady periodic frames.
  • histogram bins close to the right edge are statistically unreliable and should also be excluded from consideration when searching for peaks.
  • a periodicity histogram is computed and searched for peaks for the values of & in the predetermined interval of possible pitch periods and not for other values of k.
  • a periodicity histogram is computed and searched for peaks for the values of & in the predetermined interval of possible pitch periods and not for other values of k.
  • N 200 samples
  • m 3
  • d 10
  • the speech signal is sampled at 16 kHz.
  • a speech signal is converted into a sampled digitized format in pre-processing step 102.
  • a portion of the sampled signal (speech frame in the exemplary embodiment) is then embedded into an m-dimensional state space in step 104 to obtain a sequence of m-dimensional vectors.
  • a plurality of possible pairs of vectors in the sequence of m-dimensional vectors are considered, and the closest pairs of vectors in state space are selected in step 106.
  • a periodicity histogram is then computed in step 108 by accumulating the total number of selected closest pairs for each of the different time separation values. Then, the computed histogram is searched for highest peaks in step 110 to obtain information about pitch and periodicity.
  • the highest peak in a predetermined histogram interval is identified and its position is used to provide a pitch period estimate. More than one histogram peak can be identified and retained for use in optional subsequent post-processing step 112, which can analyze more than one consecutive frame.
  • each bin can be normalized with respect to its upper bound to produce a normalized periodicity histogram.
  • This upper bound for each bin index k is equal to the total number of vector pairs with time separation of £ samples in a set of all considered pairs of vectors.
  • EQ. 5 The difference between EQ. 5 and EQ. 4 is that the accumulated number in EQ. 5 for each value of A: is divided by the total number of pairs (M-k), so that the value of each bin cannot exceed 1. If the value of r in EQ. 5 is chosen sufficiently large, then nhist(k) - 1 for all values oik.
  • Normalized periodicity histograms obtained by normalizing the histograms of FIGS. 2B, 5B and 8B, are shown in FIGS. 2C, 5C and 8C, respectively.
  • a normalized periodicity histogram defined by EQ. 5 has a large variance at larger bin indices k approaching M due to a small number of data values involved in computing these bins.
  • the upper bound phigh of the peak-searching interval in the normalized periodicity histogram of EQ. 5 should be chosen appropriately.
  • a periodicity histogram computed according to EQ. 4 or EQ. 5, is a function of a neighborhood radius r in state space, or equivalently, of a number of selected closest pairs of vectors. The peaks in the periodicity histogram are directly affected by the value of r, or by the number of selected closest pairs of vectors in state space.
  • a space-time separation plot provides a graphical illustration of this concept: moving horizontal line 22 in FIG. 2A up or down reflects increasing or decreasing neighborhood radius, and results in more or less data points (vector pairs) located below the line and selected for computing a periodicity histogram.
  • FIGS. 3C, 6C and 9C show unbiased auto-correlation functions, computed for the same speech frames of the sustained vowel /AA/, the transitional voiced segment and the fricative /S/, respectively.
  • reconstructed frajectories for all frames are normalized to fit into the unit cube in state space, and a constant value of r is used to compute a periodicity histogram for each frame.
  • r is different for different types of signal frames.
  • an adaptive method of selecting closest pairs of vectors is used to obtain a final periodicity histogram for locating highest peaks.
  • the adaptive method which is illustrated by the flowchart in FIG. 11, can adjust a number of the selected closest pairs based on the magnitude of the highest peak in the normalized periodicity histogram.
  • the method tries to bring the highest peak's magnitude to a predetermined range of values, subject to certain constraints. Since the highest peak's magnitude is not known before the histogram is computed, the method has an iterative nature: the histogram can be recomputed several times with different numbers of selected closest pairs, each time checking the highest peak's magnitude and other conditions and adjusting the number of the selected closest pairs appropriately.
  • the adaptive method of FIG. 11 performs the following steps for each signal frame of N samples: frame 212 is embedded into an m-dimensional state space in step 214, and the resulting trajectory, described by the sequence of m-dimensional vectors, is normalized to fit into the unit cube in state space. Then, pairs of vectors closer than rmax in state space are selected from a set of possible vector pairs in the sequence of m-dimensional vectors in step 216.
  • the set of possible vector pairs includes all possible pairs of vectors with time separations between vectors in the valid search interval plow ⁇ k ⁇ phigh.
  • a normalized periodicity histogram is computed with the ntotal selected pairs in step 218, and the magnitude hmax of the highest histogram peak is determined (in the valid interval plow ⁇ k ⁇ phigh).
  • the second comparison performed in step 220 is to determine ii ntotal is less than nmin. If ntotal ⁇ nmin, then the normalized histogram from step 218 is used as the final histogram 230 without performing further steps.
  • a constant predetermined number nmin defines a minimal allowed number of vector pairs selected for computing a periodicity histogram.
  • the value of nmin is chosen to guarantee that the histogram peaks are always statistically reliable.
  • n is set equal to nmin
  • n closest pairs of vectors are selected from the set of ntotal pairs obtained in step 216. Selecting n closest pairs from the set of ntotal pairs is accomplished by ordering (sorting) the set of ntotal vector pairs by a distance in state space to form an ordered set of vector pairs, and selecting n closest pairs from this ordered set. Then, a normalized periodicity histogram is computed with the n selected closest pairs and the magnitude hmax of the highest histogram peak (in the valid histogram interval plow ⁇ k ⁇ phigh) is determined in step 224.
  • step 228 the value of n is increased and n closest pairs are selected from the set of ntotal pairs.
  • the new value of n must be less than ntotal.
  • a normalized periodicity histogram is re-computed in step 224 with the new set of n selected closest pairs.
  • the process is stopped here and the obtained normalized periodicity histogram is output as the final histogram 230.
  • the iteration loop 232 can be repeated several times, or until the condition 226 is satisfied.
  • the number of the selected closest pairs n is increased, the normalized histogram is re-computed with the new number of selected closest pairs, and the highest peak's magnitude hmax is compared to hO.
  • the final normalized periodicity histogram 230 is used for identifying highest peaks and determining pitch.
  • the computed periodicity histogram is searched for highest peaks, e.g., largest local maximums, in order to determine a fundamental period of a signal.
  • the periodicity histogram of EQ. 4 is used to identify the highest peak (the largest maximum) in the predetermined interval of possible pitch period values plow ⁇ k ⁇ phigh.
  • the peak-searching interval between plow and phigh should exclude the regions close to both left and right histogram edges.
  • the position of the identified highest peak, given by its corresponding value ofk, represents the pitch period value in samples.
  • the normalized periodicity histogram of EQ. 5 is used to identify one or more highest peaks.
  • the magnitude hmax of the highest peak in the search interval plow ⁇ k ⁇ phigh is determined.
  • fr 0.5, so that the threshold level is set at the half of the highest peak's magnitude.
  • all histogram peaks, or local maximums, with their magnitudes exceeding the threshold level thld are identified.
  • the positions and, in some embodiments, magnitudes of the identified peaks can be retained for further analysis.
  • FIG. 6A illustrates application of the above-described method of identifying highest histogram peaks as applied to the normalized periodicity histogram computed for a transitional voiced speech segment.
  • Vertical lines 61 and 62 define the lower bound plow and the upper bound phigh, respectively, of the pitch search interval.
  • the highest peak 65 inside this search interval is identified first, and the threshold level 63 is set at the fraction of the highest peak's magnitude. Then, all local peaks higher than the threshold level 63 are identified.
  • peaks 66, 67 and 68 are found to be higher than the threshold level.
  • the positions of the identified highest peaks 65, 66, 68 and 67 can be used as pitch period candidates in a post-processing stage.
  • a post-processing technique can be employed to determine a final sequence of pitch values and/or to determme whether each particular frame is periodic (voiced) or aperiodic (unvoiced).
  • the method of the present invention can produce reliable pitch estimates for clean and steady periodic frames, some form of postprocessing is usually desirable for real speech signals.
  • Post-processing allows more reliable pitch determination for frames with less than perfect periodicity, for example, fransitional or noisy speech frames. Post-processing can also be useful when one desires to reliably determine voicing state transitions in speech signals.
  • Post-processing can include analyzing positions and/or magnitudes of the identified histogram peaks for each individual frame.
  • Post-processing can also include analyzing identified histogram peaks in a larger temporal context by taking more than one consecutive frame into account.
  • the actual type of post-processing employed for a given application will, to some extent, be a function of the application's requirements.
  • the maximal allowed processing delay is a critical factor for many real-time speech-processing applications, like speech-coding devices.
  • Various different post-processing methods can also be used with the method of the present invention. For example, one can determine a final pitch value for each frame independently of other frames and, then, apply a median-smoothing technique to the obtained sequence of pitch values, in order to filter out possible incorrect values.
  • One of the most successful and popular approaches to the joint determination of pitch and voicing parameters is dynamic programming.
  • the dynamic-programming algorithm used in conjunction with the known correlation- based pitch-estimation procedure, utilizes positions and magnitudes of the highest peaks in the correlation function, in order to determine an optimal pitch frack and, at the same time, to detect voicing state transitions (Talkin, D., "A robust algorithm for pitch tracking (RAPT)", in Speech Coding and Synthesis, Elsevier, 1995, pp. 495-518).
  • Dynamic programming can and in various embodiments does, serve as the basis for a variety of different possible post-processing methods used with the present invention.
  • One feature of the present invention is directed to a simple and efficient postprocessing method, which involves simultaneous pitch tracking and voiced/unvoiced segmentation of speech signals with a minimal processing delay.
  • the highest peaks identified in the normalized periodicity histogram usually include only peaks corresponding to a fundamental pitch period and its integer multiples.
  • Such frames characterized by a high degree of periodicity, are immediately classified as voiced frames in some embodiments of the present invention.
  • the located peak positions (in number of samples) for such periodic frames are approximately related to each other as small integers 1, 2, 3 etc.
  • the pitch period value is then given by the position of the peak corresponding to 1 (the lowest sub-multiple).
  • other frames, characterized by less than perfect periodicity like the fransitional voiced frame in FIG.
  • the identified histogram peaks can also include secondary peaks caused by speech formants, and the located peak positions can deviate significantly from a simple sequence of the integer multiples of some number. For such frames, pitch can be determined more reliably by analyzing available information in a larger temporal context, that is, by examining past and future frames. The availability of the information about future frames to the pitch-tracking procedure assumes that a final decision about pitch and voicing is delayed by one or more frames.
  • each speech frame is characterized as either reliable or unreliable.
  • Speech frame is defined to be reliable if the positions of all identified highest peaks in the normalized periodicity histogram form a simple arithmetic series, like 1, 2, 3 etc.
  • a reliable speech frame can also be included in the definition of a reliable speech frame.
  • the energy of a reliable frame must exceed some predetermined threshold value.
  • the energy threshold is not a rigid value and may need to be properly adjusted in each particular case.
  • Another condition, which can be included in the definition of a reliable frame is the minimal allowed magnitude hmin of the highest peak in the normalized periodicity histogram computed with an appropriately selected neighborhood radius r. The optimal value of hmin in this case is dependent upon how the radius r is selected.
  • a frame If a frame satisfies the above conditions, it is determined to be reliable. If the above conditions are not satisfied, the frame is determined to be unreliable. A binary reliable/unreliable decision is made for each successive frame and stored for a subsequent use by a pitch-tracking procedure.
  • the steps of a pitch-tracking method implemented in accordance with one embodiment of the invention are shown in the flowchart of FIG. 12.
  • the method determines a final sequence of pitch values and classifies each frame as either voiced or unvoiced.
  • a final pitch value is assigned to each voiced frame.
  • a zero value is assigned to each unvoiced frame.
  • the method operates with a minimal delay of one frame.
  • information about the next frame Q ' +l is required by the pitch tracking method.
  • the flowchart of FIG. 12 describes pitch and voicing analysis cycle for frame j.
  • frame 0+1 is processed in step 302.
  • Processing frame 0+1) includes computing a normalized periodicity histogram and identifying highest histogram peaks.
  • a determination is made whether frame 0+1) is reliable or not.
  • a binary reliable/unreliable decision for frame 0+1) is stored for further processing. If frame 0+1) is reliable, then the located positions of all identified histogram peaks are stored as pitch period candidates in increasing order of their values (in number of samples).
  • npmax 10.
  • the analysis of frame j begins at step 304 by checking whether frame j is reliable or not. This information is available from the previous analysis cycle, when the frame index j was less by one. If frame j is reliable, then the next check is performed in step 306 whether frame 0-1) is voiced or unvoiced. The pitch period value and voicing state for frame -1) are available from the previous cycle. If frame 0-1) is voiced, then the check is performed in step 308 whether the lowest pitch period candidate of frame j matches the pitch period value of frame O ' -l). In this description of the pitch-tracking method, two pitch period values are determined to match and are classified as "matching" if their absolute difference is less than some predetermined value pdiff.
  • step 308 If the check in step 308 is positive, the decision is made in step 310 to proceed to a final step 312. In the final step 312 frame j is declared voiced and the lowest pitch period candidate of frame j becomes its final determined pitch period value. If the check in step 308 is negative, the decision is made in step 310 to proceed to step 314. In step 314 a check is performed whether the future frame 0+1) is reliable and matches frame j.
  • step 314 If frame 0+1) ⁇ s found reliable, then its lowest pitch candidate is compared to the lowest pitch candidate of frame j to determine if they match. If the check in step 314 is positive, the decision is made in step 316 to proceed to the final step 312. If the check in step 314 is negative, the decision is made in step 316 to proceed to a final step 318.
  • step 318 frame j is declared unvoiced and is assigned a zero value for the pitch period. It should be noted at this point that an unvoiced decision for frame j can be changed to voiced later by performing a backward- tracking operation in future analysis cycles.
  • step 320 a "start of voicing" check is performed.
  • the start of voicing condition is determined when two consecutive reliable frames are detected after an unvoiced frame, provided that the lowest pitch candidates for the two reliable frames match. Accordingly, the future frame 0+1) is checked in step 320 to see if it is reliable and if the lowest pitch period candidates for frames j and 0+1) match. If the start of voicing check in step 320 is positive, the decision is made in step 322 to proceed to step 324. In step 324 frame j is declared voiced and the lowest pitch period candidate becomes its final pitch period value.
  • a backward-tracking procedure is initiated in step 326.
  • the backward-tracking procedure attempts to continue pitch tracking from the current voiced frame j to past frames 0"1) > Q-2) and so on, which were previously determined to be unvoiced.
  • pitch candidates of frame 0-1) are searched for best match to the current pitch value of frame j. If the found best match does not differ from the current pitch value by more than pdiff, then frame Q ' -l) is declared voiced and the found best-matching candidate becomes the final pitch period value for frame ' -l)-
  • This backward-searching operation can be repeated for frames 0-2), (j-3) and so on, until no good match can be found.
  • the maximal allowed processing delay puts a limit on the number of frames to be considered in the backward-searching operation.
  • step 320 If the start of voicing check in step 320 is negative, the decision is made in step 322 to proceed to the final step 318.
  • step 328 determines whether frame 0-1) is voiced or unvoiced. If frame 0-1) is determined to be voiced, a forward-searching operation is performed in step 330: pitch period candidates of frame j are searched for best match to the pitch period value of the previous frame 0- )- If the found best- matching candidate does not differ from the previous pitch period value by more than pdiff, then the decision is made in step 332 to go to a final step 334. In step 334 frame j is declared voiced and the found best-matching pitch candidate becomes the final pitch period value. If no good match can be found in step 330, the decision is made in step 332 to go to the final step 318.
  • frame index j is incremented by one, and the cycle is started again. Since the analysis cycle for frame j needs information about the previously determined pitch period and voicing state for frame 0-1), the very first frame in the sequence can be initially declared unvoiced and assigned a zero for its pitch period value.
  • the obtained pitch period values can be converted into fundamental frequency values.
  • Fundamental frequency or F0, is defined as the inverse of a fundamental pitch period.
  • fundamental frequency is assigned a zero value.
  • a lookup table can be used to convert between pitch period values and fundamental frequency values.
  • FIG. 13A shows speech signal waveform of the male-spoken utterance "She had your dark suit” sampled at 16 kHz.
  • FIG. 13B shows a corresponding output of the pitch-tracking method, where each dot represents a fundamental frequency value for an individual speech frame.
  • the obtained F0 tracks may need to be further smoothed by applying some form of smoothing or best-fitting operation to successive pitch values. Such processing is contemplated and within the scope of the invention.
  • the embedding procedure used in the exemplary embodiment of the invention is time-delay embedding.
  • Time-delay embedding (or the method of delays, as it is called elsewhere) is the most widely used, but not the only known method of transforming a scalar one- dimensional signal into a frajectory in multi-dimensional space.
  • Other embedding procedures can be used, in accordance with the invention, in place of time-delay embedding to reconstruct a state-space trajectory, as long as topological properties of the original state space of a system are preserved. This means, in particular, that the reconstructed trajectory of a periodic signal should repeat itself after a complete period.
  • SVD singular value decomposition
  • the frame is first embedded using time-delay embedding with the delay parameter d and the embedding dimension of P (A DC-component should be removed prior to embedding by subtracting a mean signal value).
  • the resulting trajectory matrix X has P columns and N-(P- ⁇ )d rows:
  • S is a diagonal matrix containing singular values p ⁇ >p2 >p3 ... > 0 in decreasing order;
  • U and V are orthogonal matrices containing corresponding singular vectors.
  • the first m columns of V corresponding to largest singular values are selected and stored in V .
  • the reduced trajectory mafrix X r is obtained as follows:
  • SVD-embedding instead of time-delay embedding can be advantageous for noisy signals and some particular types of speech sounds (e.g. voiced fricatives) because of its smoothing capabilities. Smooth frajectories in state space result in a smooth periodicity histogram and, as a consequence, in better peak discrimination. However, in many cases a smoothing effect can be achieved without using SVD-embedding, by simply performing low- pass filtering of an input signal prior to its time-delay embedding.
  • the method of the present invention can produce valid results even without embedding a signal into a multi-dimensional state space. This is because the multi-dimensional embedding of a scalar signal does not contain more information than the signal itself.
  • a periodicity histogram can be computed based on absolute differences between pairs of samples, instead of distances between pairs of vectors in state space:
  • the method of the present invention remains valid when the embedding dimension m becomes equal to one, and to define one-dimensional embedding as a trivial fransformation of a signal to itself.
  • signal samples play the role of m-dimensional vectors, and that Euclidean distances in state space turn into absolute differences between sample values.
  • the number of possible pairs may be reduced to include only pairs with time separations in the predetermined interval of possible pitch periods.
  • the procedure of generating all possible non-repeating pairs of vectors which corresponds to the definition of a periodicity histogram in EQ. 4, can be better understood using the schematic illustration in FIG. 15A.
  • the decreasing number of pairs leads to a fall-off of the histogram peaks as k is increased, and causes the histogram bins with larger indices k near M to be statistically unreliable.
  • Normalizing each bin of the periodicity histogram in accordance with EQ. 5 prevents a fall-off of peak magnitudes at larger k, but does not make all bins statistically equally reliable.
  • the normalized histogram of EQ. 5 has a large variance at larger bin indices k approaching M, as the number of pairs involved in computing corresponding histogram bins becomes small. It is interesting to observe some analogy here with the conventional definitions of biased and unbiased auto-correlation functions.
  • the procedure of generating this subset of pairs can be better understood using the schematic illustration in FIG. 15B.
  • the upper row of dots 156 represents the sequence of vectors x(t) (i- ⁇ ..M).
  • the lower row of dots 158 represents a subsequence of the sequence 156.
  • the summation interval is the same for all k, so that an equal number of pairs is involved in calculating each bin value. All histogram peaks are thus normalized with respect to the same constant number and are equally reliable statistically.
  • the modified periodicity histogram is used in place of the normalized periodicity histogram in one embodiment of the invention.
  • the peak-searching interval in the modified histogram can be extended to the right edge, since all histogram bins are now equally reliable.
  • the peaks in the periodicity histogram are usually much sharper and can have a rough appearance in many cases. This can be observed, for example, in FIGS. 5C, 6A and 6B.
  • the rough appearance can cause undesirable effects in some cases when histogram peaks are identified, especially with noisy signals.
  • additional local maxima can sometimes be detected in the vicinity of an identified large peak. Therefore, in order to facilitate peak discrimination, it can be advantageous to obtain a smoothed histogram before searching for local peaks.
  • One way to obtain a smoothed periodicity histogram is to start with a smooth trajectory in m-dimensional state-space, provided the employed sampling rate is sufficient. Smooth trajectory can be obtained by performing low-pass filtering of the input signal before embedding it. Alternatively, SVD-embedding procedure can be used with an appropriately chosen SVD-window length.
  • the histogram can be smoothed using any of the conventional smoothing methods.
  • a simple 3 -point moving-average smoothing procedure is used for this purpose.
  • any suitable smoothing or curve-fitting procedure can be applied to a histogram, in order to achieve more reliable peak discrimination.
  • An alternative approach is to apply some averaging operation to a distribution of spatio-temporal distances in the r direction. For example, a periodicity histogram can be computed several times, each time changing the value of r by some ⁇ r. Then, a weighted average of these computed histograms can be used as a final smooth histogram for peak searching:
  • ⁇ nalhist(k) w ⁇ *nhist(k, r- ⁇ r) + w2*nhist(k, r) + w3*nhist(k, r+ ⁇ r) EQ. 12
  • the method of the present invention involves selecting closest pairs of vectors from a set of possible vector pairs formed in the sequence of M vectors in m-dimensional state space.
  • M is the number of m-dimensional vectors obtained after embedding a signal frame.
  • the value of Mis proportional to a sampling rate and to a frame size and is typically a few hundred.
  • Finding nearest-neighbor points in multi-dimensional space is an extensively studied subject in computational geometry. Nearest-neighbor search is also one of the frequently encountered tasks in nonlinear and chaotic time-series analysis (e.g. Schreiber, T., "Efficient neighbor searching in nonlinear time series analysis", Int. J. Bifurcation and Chaos, 5, 1995, p. 349).
  • a number of fast neighbor-searching algorithms have been developed to date.
  • the two most popular approaches, described in the literature, are free-based search methods and box-assisted search methods.
  • any suitable algorithm can be used in connection with the present invention, the selection of best-performing algorithm depends on many factors, such as signal properties, embedding dimension, sampling rate etc. For example, with low sampling rate and/or small number of samples in a frame, the value of Mis small, and a simple computation of all distances may actually be cheaper than using a sophisticated fast algorithm.
  • Another effective method of reducing computational cost is to compute a periodicity histogram using a down-sampled version of a signal first. This down-sampled version of a histogram is searched for highest peaks in the full pitch search range (between plow and phigh search bounds). After the highest peaks are identified, the histogram is computed at the original sampling rate, but only in the vicinity of the identified highest peaks. The peak positions are then determined more accurately.
  • the present invention provides a reliable, accurate and efficient method for determining pitch and/or periodicity of speech signals.
  • the invention also provides an efficient method for pitch tracking and/or for performing segmentation of speech signals into voiced and unvoiced portions.
  • a pitch period value may be generated.
  • a pitch period value is to be interpreted as a value that is indicative of the fundamental period of a signal or a portion of a signal.
  • the invention can be implemented in software, hardware, or any combination of software and hardware.
  • FIG. 16 illusfrates a schematic block diagram of a pitch determination apparatus 1600 in the form of a digital signal processor 1602 used in conjunction with an analog to digital converter 1604, which can also include other parts and can itself be included in any device.
  • the digital signal processor 1602 may be used as a pitch detector in a speech-coding device, a speech recognition system, a speaker recognition system and a speech synthesis system.
  • the digital signal processor 1602 includes a CPU 1608 for executing instructions included in the software of the present invention.
  • the software is stored in program instructions memory 1606.
  • the digital signal processor 1602 receives digitized speech from the A/D converter 1604, processes it in accordance with the present invention, and outputs a resulting pitch signal which assumes a value indicative of the detected pitch of the speech signal at a particular point in time.
  • the CPU 1608 may use data memory 1610 to store samples, vectors and/or other values used as part of the pitch determination method of the present invention.
  • the invention can be embodied in a set of machine readable instructions stored on a digital data storage device such as a RAM, ROM or disk type of storage.
  • a digital data storage device such as a RAM, ROM or disk type of storage.
  • the machine readable instructions in the software of the invention confrol a processor and/or other hardware to perform the steps of the present invention.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Abstract

Procédés et appareil de détection de la périodicité et/ou de détermination de la période fondamentale d'un signal tel qu'un signal de communication vocale. Les procédés selon l'invention consistent (104) à intégrer une portion d'un signal numérisé échantillonné dans un espace d'états à m dimensions, de manière à obtenir une séquence de m vecteurs dimensionnels, (106) à sélectionner les paires les plus rapprochées de vecteurs dans l'espace d'états, à partir d'une pluralité de paires possibles de m vecteurs dimensionnels dans ladite séquence de vecteurs, (108) à cumuler les nombres totaux de paires rapprochées sélectionnées de vecteurs ayant les mêmes valeurs de séparation temporelle, de manière à obtenir un histogramme de nombres cumulés et (110), à repérer au moins un pic maximum dans une portion dudit histogramme, de manière à obtenir une valeur indiquant la période fondamentale du signal. Différentes formes d'exécution concernent le traitement du signal vocal et audio et autres applications ayant trait à la communication vocale. Les procédés ont toutefois un caractère général et peuvent être également appliqués à d'autres types de signaux périodiques ou quasi-périodiques.
PCT/US2002/033895 2001-10-26 2002-10-23 Procedes et appareil de determination de la hauteur tonale WO2003038806A1 (fr)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
US34888301P 2001-10-26 2001-10-26
US60/348,883 2001-10-26
US10/140,211 2002-05-07
US10/140,211 US7124075B2 (en) 2001-10-26 2002-05-07 Methods and apparatus for pitch determination

Publications (1)

Publication Number Publication Date
WO2003038806A1 true WO2003038806A1 (fr) 2003-05-08

Family

ID=26837975

Family Applications (2)

Application Number Title Priority Date Filing Date
PCT/US2002/032987 WO2003038805A1 (fr) 2001-10-26 2002-10-16 Procedes et appareil permettant de determiner une hauteur tonale
PCT/US2002/033895 WO2003038806A1 (fr) 2001-10-26 2002-10-23 Procedes et appareil de determination de la hauteur tonale

Family Applications Before (1)

Application Number Title Priority Date Filing Date
PCT/US2002/032987 WO2003038805A1 (fr) 2001-10-26 2002-10-16 Procedes et appareil permettant de determiner une hauteur tonale

Country Status (3)

Country Link
US (1) US7124075B2 (fr)
EP (1) EP1451804A4 (fr)
WO (2) WO2003038805A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7740618B2 (en) 2004-10-14 2010-06-22 Novo Nordisk A/S Syringe device

Families Citing this family (30)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7529661B2 (en) * 2002-02-06 2009-05-05 Broadcom Corporation Pitch extraction methods and systems for speech coding using quadratically-interpolated and filtered peaks for multiple time lag extraction
US7752037B2 (en) * 2002-02-06 2010-07-06 Broadcom Corporation Pitch extraction methods and systems for speech coding using sub-multiple time lag extraction
US7236927B2 (en) * 2002-02-06 2007-06-26 Broadcom Corporation Pitch extraction methods and systems for speech coding using interpolation techniques
KR100596099B1 (ko) * 2002-11-11 2006-07-06 독립행정법인전자항법연구소 심신 진단 시스템
US7352373B2 (en) * 2003-09-30 2008-04-01 Sharp Laboratories Of America, Inc. Systems and methods for multi-dimensional dither structure creation and application
US7386536B1 (en) * 2003-12-31 2008-06-10 Teradata Us, Inc. Statistical representation of skewed data
DE102004045097B3 (de) * 2004-09-17 2006-05-04 Carl Von Ossietzky Universität Oldenburg Verfahren zur Extraktion periodischer Signalkomponenten und Vorrichtung hierzu
US7933767B2 (en) * 2004-12-27 2011-04-26 Nokia Corporation Systems and methods for determining pitch lag for a current frame of information
US8738370B2 (en) * 2005-06-09 2014-05-27 Agi Inc. Speech analyzer detecting pitch frequency, speech analyzing method, and speech analyzing program
KR100653643B1 (ko) * 2006-01-26 2006-12-05 삼성전자주식회사 하모닉과 비하모닉의 비율을 이용한 피치 검출 방법 및피치 검출 장치
US7805308B2 (en) * 2007-01-19 2010-09-28 Microsoft Corporation Hidden trajectory modeling with differential cepstra for speech recognition
ATE504010T1 (de) * 2007-06-01 2011-04-15 Univ Graz Tech Gemeinsame positions-tonhöhenschätzung akustischer quellen zu ihrer verfolgung und trennung
EP2162880B1 (fr) * 2007-06-22 2014-12-24 VoiceAge Corporation Procédé et dispositif d'estimation de la tonalité d'un signal sonore
DE102007030209A1 (de) * 2007-06-27 2009-01-08 Siemens Audiologische Technik Gmbh Glättungsverfahren
CA2657087A1 (fr) * 2008-03-06 2009-09-06 David N. Fernandes Systeme de base de donnees et methode applicable
US8380331B1 (en) 2008-10-30 2013-02-19 Adobe Systems Incorporated Method and apparatus for relative pitch tracking of multiple arbitrary sounds
CN101604525B (zh) * 2008-12-31 2011-04-06 华为技术有限公司 基音增益获取方法、装置及编码器、解码器
WO2010091554A1 (fr) * 2009-02-13 2010-08-19 华为技术有限公司 Procédé et dispositif de détection de période de pas
US8515196B1 (en) * 2009-07-31 2013-08-20 Flir Systems, Inc. Systems and methods for processing infrared images
JP5177157B2 (ja) * 2010-03-17 2013-04-03 カシオ計算機株式会社 波形発生装置および波形発生プログラム
US20130080165A1 (en) * 2011-09-24 2013-03-28 Microsoft Corporation Model Based Online Normalization of Feature Distribution for Noise Robust Speech Recognition
US8965832B2 (en) 2012-02-29 2015-02-24 Adobe Systems Incorporated Feature estimation in sound sources
US8949118B2 (en) * 2012-03-19 2015-02-03 Vocalzoom Systems Ltd. System and method for robust estimation and tracking the fundamental frequency of pseudo periodic signals in the presence of noise
US20130282373A1 (en) * 2012-04-23 2013-10-24 Qualcomm Incorporated Systems and methods for audio signal processing
US9589570B2 (en) 2012-09-18 2017-03-07 Huawei Technologies Co., Ltd. Audio classification based on perceptual quality for low or medium bit rates
JP5995226B2 (ja) * 2014-11-27 2016-09-21 インターナショナル・ビジネス・マシーンズ・コーポレーションＩｎｔｅｒｎａｔｉｏｎａｌ Ｂｕｓｉｎｅｓｓ Ｍａｃｈｉｎｅｓ Ｃｏｒｐｏｒａｔｉｏｎ 音響モデルを改善する方法、並びに、音響モデルを改善する為のコンピュータ及びそのコンピュータ・プログラム
US9842611B2 (en) * 2015-02-06 2017-12-12 Knuedge Incorporated Estimating pitch using peak-to-peak distances
CN104794175B (zh) * 2015-04-01 2018-01-23 浙江大学 基于度量k最近对的景点和酒店最佳配对方法
US10283143B2 (en) 2016-04-08 2019-05-07 Friday Harbor Llc Estimating pitch of harmonic signals
US10229092B2 (en) * 2017-08-14 2019-03-12 City University Of Hong Kong Systems and methods for robust low-rank matrix approximation

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3649765A (en) * 1969-10-29 1972-03-14 Bell Telephone Labor Inc Speech analyzer-synthesizer system employing improved formant extractor
US3916105A (en) * 1972-12-04 1975-10-28 Ibm Pitch peak detection using linear prediction
US6216118B1 (en) * 1996-10-31 2001-04-10 Kabushiki Kaisha Meidensha Apparatus and method for discriminating a time series data

Family Cites Families (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US2908761A (en) * 1954-10-20 1959-10-13 Bell Telephone Labor Inc Voice pitch determination
US3405237A (en) * 1965-06-01 1968-10-08 Bell Telephone Labor Inc Apparatus for determining the periodicity and aperiodicity of a complex wave
US3496465A (en) * 1967-05-19 1970-02-17 Bell Telephone Labor Inc Fundamental frequency detector
US3535454A (en) * 1968-03-05 1970-10-20 Bell Telephone Labor Inc Fundamental frequency detector
US3566035A (en) * 1969-07-17 1971-02-23 Bell Telephone Labor Inc Real time cepstrum analyzer
US3740476A (en) * 1971-07-09 1973-06-19 Bell Telephone Labor Inc Speech signal pitch detector using prediction error data
US4015088A (en) * 1975-10-31 1977-03-29 Bell Telephone Laboratories, Incorporated Real-time speech analyzer
JPS58140798A (ja) * 1982-02-15 1983-08-20 株式会社日立製作所 音声ピツチ抽出方法
US4672667A (en) * 1983-06-02 1987-06-09 Scott Instruments Company Method for signal processing
US4879748A (en) * 1985-08-28 1989-11-07 American Telephone And Telegraph Company Parallel processing pitch detector
US5226108A (en) * 1990-09-20 1993-07-06 Digital Voice Systems, Inc. Processing a speech signal with estimated pitch
US5751905A (en) * 1995-03-15 1998-05-12 International Business Machines Corporation Statistical acoustic processing method and apparatus for speech recognition using a toned phoneme system
WO1997027578A1 (fr) * 1996-01-26 1997-07-31 Motorola Inc. Analyseur de la parole dans le domaine temporel a tres faible debit binaire pour des messages vocaux
US6047254A (en) * 1996-05-15 2000-04-04 Advanced Micro Devices, Inc. System and method for determining a first formant analysis filter and prefiltering a speech signal for improved pitch estimation
US6026357A (en) * 1996-05-15 2000-02-15 Advanced Micro Devices, Inc. First formant location determination and removal from speech correlation information for pitch detection
FI113903B (fi) * 1997-05-07 2004-06-30 Nokia Corp Puheen koodaus
US5960387A (en) * 1997-06-12 1999-09-28 Motorola, Inc. Method and apparatus for compressing and decompressing a voice message in a voice messaging system
KR100269216B1 (ko) * 1998-04-16 2000-10-16 윤종용 스펙트로-템포럴 자기상관을 사용한 피치결정시스템 및 방법
US6226606B1 (en) * 1998-11-24 2001-05-01 Microsoft Corporation Method and apparatus for pitch tracking
DE19859174C1 (de) * 1998-12-21 2000-05-04 Max Planck Gesellschaft Verfahren und Vorrichtung zur Verarbeitung rauschbehafteter Schallsignale
US6584437B2 (en) * 2001-06-11 2003-06-24 Nokia Mobile Phones Ltd. Method and apparatus for coding successive pitch periods in speech signal

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US3649765A (en) * 1969-10-29 1972-03-14 Bell Telephone Labor Inc Speech analyzer-synthesizer system employing improved formant extractor
US3916105A (en) * 1972-12-04 1975-10-28 Ibm Pitch peak detection using linear prediction
US6216118B1 (en) * 1996-10-31 2001-04-10 Kabushiki Kaisha Meidensha Apparatus and method for discriminating a time series data

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7740618B2 (en) 2004-10-14 2010-06-22 Novo Nordisk A/S Syringe device

Also Published As

Publication number Publication date
US20030088401A1 (en) 2003-05-08
EP1451804A4 (fr) 2005-11-23
EP1451804A1 (fr) 2004-09-01
WO2003038805A1 (fr) 2003-05-08
US7124075B2 (en) 2006-10-17

Similar Documents

Publication Publication Date Title
US7124075B2 (en) Methods and apparatus for pitch determination
KR100873396B1 (ko) 오디토리 이벤트에 기초한 특성을 이용하여 오디오를비교하는 방법
US7711123B2 (en) Segmenting audio signals into auditory events
Ying et al. A probabilistic approach to AMDF pitch detection
JP2004528599A (ja) オーディトリーイベントに基づく特徴付けを使ったオーディオの比較
US8942977B2 (en) System and method for speech recognition using pitch-synchronous spectral parameters
US20090089051A1 (en) Vocal fry detecting apparatus
CN108847218B (zh) 一种自适应门限整定语音端点检测方法，设备及可读存储介质
US7966179B2 (en) Method and apparatus for detecting voice region
US7043424B2 (en) Pitch mark determination using a fundamental frequency based adaptable filter
JPH06161494A (ja) 音声のピッチ区間自動抽出方法
Sripriya et al. Pitch estimation using harmonic product spectrum derived from DCT
Ziólko et al. Phoneme segmentation of speech
Zhao et al. A processing method for pitch smoothing based on autocorrelation and cepstral F0 detection approaches
US4982433A (en) Speech analysis method
Upadhya Pitch detection in time and frequency domain
KR0136608B1 (ko) 음성신호 검색용 음성인식 장치
Arroabarren et al. Glottal source parameterization: a comparative study
JPH01255000A (ja) 音声認識システムに使用されるテンプレートに雑音を選択的に付加するための装置及び方法
KR100194953B1 (ko) 유성음 구간에서 프레임별 피치 검출 방법
Tantisatirapong et al. Comparison of feature extraction for accent dependent Thai speech recognition system
Kuberski et al. A landmark-based approach to automatic voice onset time estimation in stop-vowel sequences
Bonifaco et al. Comparative analysis of filipino-based rhinolalia aperta speech using mel frequency cepstral analysis and Perceptual Linear Prediction
Wiriyarattanakul et al. A Syllable-based Speech Recognition system by using Pitch detection on Time-Frequency domain Feature Extraction
Chaari et al. Wavelet ridge track interpretation in terms of formants

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ OM PH PL PT RO RU SD SE SG SI SK SL TJ TM TN TR TT TZ UA UG US UZ VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR IE IT LU MC NL PT SE SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
122 Ep: pct application non-entry in european phase
NENP Non-entry into the national phase

Ref country code: JP

WWW Wipo information: withdrawn in national office

Country of ref document: JP